Collin Francel
(615) 239-3400 | collin.t.francel@proton.me | github.com/xXCoolinXx
| University of Alabama: Graduate | Tuscaloosa, AL |
| Master of Arts in Mathematics | Fall 2025 – Fall 2026 |
| University of Alabama: Undergraduate | Tuscaloosa, AL |
| Bachelor of Science in Data Science, Minor in Spanish, Randall Research Scholar, GPA: 4.0 | Aug. 2023 – Fall 2026 |
| Undergraduate Researcher | Aug. 2024 – May 2025 |
| University of Alabama | Tuscaloosa, AL |
Worked with STANDD Initiative, an initiative to use data analysis techniques to fight human trafficking
Used language analysis techniques and encoder language models to analyze text data in illicit ads
Developed method with improvement of 5-10% in accuracy and 10-15% in recall over previous technique
| ORISE Undergraduate Research Intern | June 2024 – Aug. 2024 |
| Oak Ridge National Lab | Oak Ridge, TN |
Explored novel approach combining language models and graph neural networks for modeling organic molecules
Designed and executed large scale hyperparameter optimization on Frontier supercomputer with 100+ trials across 8 nodes per method, ensuring robust and fair comparison between approaches
Contributed valuable insights on method limitations, leading to accepted NeurIPS workshop paper
CollinGPT-2 | Python, Cloud Computing, Qwen, transformers, llama.cpp |
May 2025 |
Developed chat-bot fine-tuned on my personal writing and essays to emulate my personality in conversation
Synthesized and cleaned dataset of over 30 million tokens of my personal writing for training
Employed cloud computing resources to efficiently fine-tune the Qwen3-8B model in just 15 minutes
Utilized the llama.cpp program to quantize the model for local inference on consumer hardware
| SQLLM | Python, NiceGUI, HuggingFace, CodeLlama, Git | February 2024 |
Created an AI-powered interface enabling non-technical users to query databases using natural language
Integrated CodeLlama-34B to accurately translate user requests into optimized SQL commands
Designed and implemented custom prompt engineering techniques to ensure robust SQL generation
Built intuitive front-end with NiceGUI, resulting in seamless database interaction for users with no SQL knowledge
Relevant Coursework: Earned A+ in courses including Data Structures and Algorithms, Software Engineering, Database Management Systems, Theory of Probability, Nonlinear Optimization Theory, and Math Stats w/ Applications
Languages: Python, Bash, C++, C, Web (HTML, CSS, Javascript), SQL
Developer Tools: Slurm, Conda, Git, Github, HuggingFace, VS Code, Jupyter Notebooks
Libraries: pandas/polars, NumPy, Matplotlib, scipy, scikit-learn, pytorch, Avalanche
Languages: English - Native, Spanish - Intermediate/Advanced